By 2030, data centers will require $6.7 trillion worldwide. That number — from McKinsey’s most rigorous technology infrastructure analysis to date — represents roughly the combined GDP of Japan and Germany. The AI data center cycle is regularly compared to the 1990s fiber overbuild. It is a seductive analogy and a fundamentally misleading one. Understanding why is the difference between capturing a decade-long compounding trade and being burned by a narrative that looked good on paper.

McKinsey’s research shows global demand for data center capacity could almost triple by 2030, with approximately 70% of that demand driven by AI workloads. Total projected capital expenditure: $6.7 trillion, of which $5.2 trillion is attributable to AI processing loads and $1.5 trillion to traditional IT applications.

The hyperscalers are leading the investment wave. Amazon, Google, Microsoft, and Meta are expected to spend over $350 billion on capex in 2025 alone — a year-over-year increase in the mid-30% range. In aggregate, AI-related infrastructure spend in 2025 is estimated at approximately $500 billion, and in H1 2025 it contributed more to US GDP growth than consumer spending. As a share of GDP, AI-related capex now sits at approximately 5% — a level comparable to the late-1990s technology boom.

Three Investment Scenarios

McKinsey constructed three scenarios ranging from constrained to accelerated demand, shaped by semiconductor supply constraints, enterprise AI adoption rates, efficiency improvements, and regulatory challenges. The base case — $5.2 trillion in AI data center capex — assumes continued growth without runaway acceleration or structural constraints.

The analogy to the late-1990s telecommunications infrastructure bubble is compelling in one dimension — the scale of capital deployment — and misleading in every other. Fiber in the 1990s was built speculatively, with virtually unlimited capacity once laid and zero refresh requirement. Data centers are physically constrained, contractually committed before construction, and subject to accelerated depreciation cycles that naturally absorb any temporary excess. The evidence is visible in vacancy data: North American colocation vacancy has fallen from 9.8% in 2020 to 2.3% in H1 2025 (JLL Research), while the fiber glut post-2001 saw vacancy exceed 20%.

The key structural difference McKinsey identifies is the cost of carrying excess capacity. Fiber, once laid, is nearly free to maintain. A dark fiber network can sit idle for years without material cost. Data centers are the opposite: power, cooling, and maintenance are ongoing high costs regardless of utilization. But crucially, AI accelerators have 3–4 year refresh cycles — meaning any overcapacity is rapidly converted into obsolescence, and new workloads pull spare capacity well before it becomes stranded. This creates a natural self-correction mechanism the 1990s fiber cycle entirely lacked.

McKinsey’s analysis maps the $5.2 trillion AI capex envelope across five distinct investor archetypes. Understanding this architecture is essential: the investment case, risk profile, and return dynamics differ fundamentally across archetypes. Three archetypes receive direct quantified capex allocation; two (Operators and AI Architects) are excluded from the model because their compute investment overlaps with broader R&D spending.

Any credible investment thesis requires honest engagement with the counter-case. McKinsey explicitly identifies the critical uncertainties that could derail even the base-case scenario. The investment question is not whether these risks are real — they are — but whether they fundamentally alter the long-run structural thesis or merely create volatility in the path.

The $5.2–$6.7 trillion capex envelope flows through a defined set of public equities. But raw exposure to the AI theme is not sufficient — the archetype, moat, and balance sheet quality of each company determine whether they capture compounding returns or get crushed in the shake-out. The following analysis maps our highest-conviction positions, selective holds, and explicit avoids — with rationale grounded in the McKinsey investment framework.

High Conviction: Own the Moats

Selective Positions: Conditional on Execution

Avoid: Where the Shake-Out Will Concentrate

Based on the McKinsey demand model and our proprietary analysis, the following table maps projected asset-class and sector-level impacts by phase of the AI infrastructure cycle. The cycle has three phases: Build (2024–26, hardware-dominant), Deploy (2026–28, software and efficiency), and Compound (2028–30, productivity realisation).

The alternative asset management sector is not broken. It is repriced. That distinction — cyclical compression versus structural impairment — is the most important investment judgement available in the financial services sector in 2026. Get it right and you own a multi-year compounding trade at below-average multiples. Get it wrong and you are catching a knife into a structural earnings decline.

Force 1: The Rate Shock Transmission

Private equity is structurally leveraged. Buyout funds typically finance 50–60% of acquisition cost with debt. When the risk-free rate rose from near-zero to 5%+, the cost of leverage on portfolio companies increased materially, compressing free cash flow and triggering valuation mark-downs in parallel. Companies acquired at 2020–2022 vintage valuations using cheap floating-rate debt now face a dual squeeze: higher interest expense and de-rated public market comparables. The impact on performance-linked earnings (carried interest) has been significant — carry receipts fell sharply from 2021–22 peaks and have been slow to recover as GPs hold assets rather than crystallise losses at reduced exit multiples.

Force 2: The 2021 Vintage Problem

The worst-positioned vintage in recent PE history is 2021. Global PE deal value exceeded $1.1 trillion that year, at median buy-in EV/EBITDA multiples of approximately 11× — near the all-time high. Those assets now trade at 7–8× in comparable public markets. The challenge is not merely paper marks but the exit path: to return capital from 2021-vintage investments, GPs must either accept lower exit multiples than at entry (crystallising losses) or hold longer (extending fund lives and delaying carried interest). Neither is favourable for near-term earnings.

Force 3: Public Market Re-Rating of Earnings Quality

Alternative asset managers have been re-rated on two dimensions simultaneously: actual earnings pressure (carry receipts falling) and earnings quality perception (markets assigning lower multiples to lumpy, mark-to-market-dependent carry). The more defensible earnings stream — Fee-Related Earnings (FRE) — derives from management fees (typically 1.5–2% of committed capital) and is largely independent of market conditions and realisation timing. This stream has continued to grow at mid-to-high teens rates even as carry has fallen. The investment thesis for the sector ultimately reduces to: is current pricing giving you the FRE for free?

BX, KKR, APO, and ARES are frequently discussed as a homogeneous group. They are not. Their earnings composition, vintage exposure, structural differentiators, and trajectory under different rate scenarios differ materially — and selecting the right ones at the right entry points is where the real investment alpha lies.